[Sweclockers] Radeon 380X coming late spring, almost 50% improvement over 290X

[TurtleCrusher]

If anyone is going to whine about a high end 300w video card, we should just take a stroll through their post history. I remember tons of people on here were recommending Crossfire/SLI 5850/6870s and 460s on lower to mid-range builds. Either of those in pairs are 300w. This card is ultra high end and 300w load is not unreasonable.

It is hilarious to know those same people who complain about a 40w power consumption spread from current i7s to FX used to (and some still do) overclock first-gen i7 cpus past 4GHz. Those ate up 250-300w alone!

If any of you are that worried about power consumption, you shouldn't be worried about high-end anything. All this power consumption nonsense is just used a fud to talk smack about things people don't like. There's posters in here who rant so hard about power consumption yet drive a huge truck to work in the city for a software job... You know who you are!

 

[Magic Carpet]

If anyone is going to whine about a high end 300w video card, we should just take a stroll through their post history. I remember tons of people on here were recommending Crossfire/SLI 5850/6870s and 460s on lower to mid-range builds. Either of those in pairs are 300w. This card is ultra high end and 300w load is not unreasonable.

It is hilarious to know those same people who complain about a 40w power consumption spread from current i7s to FX used to (and some still do) overclock first-gen i7 cpus past 4GHz. Those ate up 250-300w alone!

If any of you are that worried about power consumption, you shouldn't be worried about high-end anything. All this power consumption nonsense is just used a fud to talk smack about things people don't like. There's posters in here who rant so hard about power consumption yet drive a huge truck to work in the city for a software job... You know who you are!

Well put :thumbsup:

 

[exar333]

If anyone is going to whine about a high end 300w video card, we should just take a stroll through their post history. I remember tons of people on here were recommending Crossfire/SLI 5850/6870s and 460s on lower to mid-range builds. Either of those in pairs are 300w. This card is ultra high end and 300w load is not unreasonable.

It is hilarious to know those same people who complain about a 40w power consumption spread from current i7s to FX used to (and some still do) overclock first-gen i7 cpus past 4GHz. Those ate up 250-300w alone!

If any of you are that worried about power consumption, you shouldn't be worried about high-end anything. All this power consumption nonsense is just used a fud to talk smack about things people don't like. There's posters in here who rant so hard about power consumption yet drive a huge truck to work in the city for a software job... You know who you are!

For one, that 40w spread gets fatter and fatter as you ramp-up clock-speeds. So it does matter quite a bit. Enough so that AMD had to essentially create a whole-other motherboard socket because the current one couldn't support enough voltage for higher-end CPUs. That's not 'good'. Works, but not ideal.

Second, the first-gen i7s were already the best around when they launched. They did eat-up power as you overclocked them (I had the 920, so I know and can vouch for the power-hungry nature) but performance also scaled really well, and you went from the best to even better.

AMD fans are saying the same thing that NV fans did when Fermi was out. 'Efficiency doesn't matter...' It does, actually. That's why AMD is dying on the vine for server products while Intel is cramming more and more cores onto their dies.

Efficiency mattered when A64 was killing Netburst and it still matters. Saying otherwise is just sour grapes IMHO...

 

[Creig]

AMD fans are saying the same thing that NV fans did when Fermi was out. 'Efficiency doesn't matter...' It does, actually. That's why AMD is dying on the vine for server products while Intel is cramming more and more cores onto their dies.

We're not discussing servers here. We're talking about enthusiast PCs. Power efficiency takes a back seat to Price/Performance for our computing segment.

 

[AnandThenMan]

Efficiency is ONE metric, it's important but not everything. Also efficiency didn't matter for Netburst because it outsold AMD by a huge margin, Fermi also sold extremely well. Nvidia lost market share because Fermi was late not because it was extremely power hungry.

We're not discussing servers here. We're talking about enthusiast PCs. Power efficiency takes a back seat to Price/Performance for our computing segment.

Exactly.

 

[TurtleCrusher]

For one, that 40w spread gets fatter and fatter as you ramp-up clock-speeds. So it does matter quite a bit. Enough so that AMD had to essentially create a whole-other motherboard socket because the current one couldn't support enough voltage for higher-end CPUs. That's not 'good'. Works, but not ideal.

Second, the first-gen i7s were already the best around when they launched. They did eat-up power as you overclocked them (I had the 920, so I know and can vouch for the power-hungry nature) but performance also scaled really well, and you went from the best to even better.

AMD fans are saying the same thing that NV fans did when Fermi was out. 'Efficiency doesn't matter...' It does, actually. That's why AMD is dying on the vine for server products while Intel is cramming more and more cores onto their dies.

Efficiency mattered when A64 was killing Netburst and it still matters. Saying otherwise is just sour grapes IMHO...

Serious overclocking and worry about power consumption doesn't even belong in the same conversation. If you're that much of a penny-pincher or live in a place where electricity is outrageous, you're in the wrong hobby. Simple as that.
A very select few people overclock in the first place, even with unlocked K and FX CPUS. 40w spread is nothing. Change a desk lamp to an LED light, problem solved.

 

[exar333]

Serious overclocking and worry about power consumption doesn't even belong in the same conversation. If you're that much of a penny-pincher or live in a place where electricity is outrageous, you're in the wrong hobby. Simple as that.
A very select few people overclock in the first place, even with unlocked K and FX CPUS. 40w spread is nothing. Change a desk lamp to an LED light, problem solved.

LOL. Yeah right....

Power/heat don't matter at all to an overclocker. I am pretty sure it matters a lot when you have to design cooling for all those components. A few hundred watts you claim to be 'immaterial' can hugely affect how you design and build a new rig, be it air or water-cooled.

Try designing a custom loop for 600w vs. 1000w. You get the idea...It's not the cost, its called physics.

 

[JoeRambo]

If product is power efficient, GPU manufacturers can convert that efficiency into more performance, simple as that. Or they can leave that perf on the table for customer. In a world where you can't break certain power limits to still have PCIE aproval, max perf for 300 watts is important stock perf limiter.

Still, if talks about HBM are true, 300w of pure GPU power with massive bw, reduced latency are gonna be epic!

 

[videogames101]

LOL. Yeah right....

Power/heat don't matter at all to an overclocker. I am pretty sure it matters a lot when you have to design cooling for all those components. A few hundred watts you claim to be 'immaterial' can hugely affect how you design and build a new rig, be it air or water-cooled.

Try designing a custom loop for 600w vs. 1000w. You get the idea...It's not the cost, its called physics.

The market segment containing overclocking enthusiasts is about performance above power consumption. When you're OC'ing heavily, you're at least doubling the power draw of a CPU. Why would we do that if we gave a rat's ass about power consumption?

The fact of the matter is if the product can be cooled reasonably as it appears it will be with a hybrid cooler, or as CPU's can be with water or large air towers, overclockers will happily sacrifice power consumption for speed.

Give me a faster product that consumes more power than the competition but is still cooled properly and I'll always take the faster one. Many others feel this way. That being said, in my next build I'll be trying to make ITX work and will probably not be selecting a 380X because of that fact.

 

[RampantAndroid]

The market segment containing overclocking enthusiasts is about performance above power consumption. When you're OC'ing heavily, you're at least doubling the power draw of a CPU. Why would we do that if we gave a rat's ass about power consumption?

Because it matters if a CPU is eating up 40W less (and generating less heat) at stock temps, allowing you to push the voltage and get a decent overclock before you have to resort to a custom water loop?

 

[TurtleCrusher]

LOL. Yeah right....

Power/heat don't matter at all to an overclocker. I am pretty sure it matters a lot when you have to design cooling for all those components. A few hundred watts you claim to be 'immaterial' can hugely affect how you design and build a new rig, be it air or water-cooled.

Try designing a custom loop for 600w vs. 1000w. You get the idea...It's not the cost, its called physics.

How many people do water cooling or heavily overclocked systems? The vast majority of people do not with their desktops. That's such a tiny, tiny part of the market. For practically everyone, it's literally a 80w vs 125w comparison, and that's at load. A sub $20 heatsink keeps everything cool and quiet even with a decent overclock. At idle there's barely a difference and it has more to do with individual motherboards than processors.

~40w is nothing. If you're doing serious overclocking then you plan accordingly. That's such a niche market though. It's like talking to daily commuters about the effects of adding a turbocharger to their car.

 

[RussianSensation]

LOL. Yeah right....

Power/heat don't matter at all to an overclocker. I am pretty sure it matters a lot when you have to design cooling for all those components. A few hundred watts you claim to be 'immaterial' can hugely affect how you design and build a new rig, be it air or water-cooled.

You have a 5820 @ 4.5Ghz. Why did you get that over a 4790K @ 4.5Ghz as in games it hardly matters?

1) 5820K OC platform wasted 32W+ of power at idle which at 24/7 operation adds up.

2) 5820K @ 4.3Ghz uses 76% more power than a Core i7 4790K @ 4.5Ghz system or 129W of power, MORE than the difference between a 290X and a 970!

Why did you get a 5820K over the 4790K in your gaming rig if you care so much about power? Could it possibly be that you chose performance in multi-threaded apps and future-proofing over power usage and electricity savings?

Sorry but I'll take a 300W GPU over a 200W GPU even if it was just 20-30% faster. I would also take a $252 R9 290X and 100W more power over a $350 MSI Gaming GTX970 cuz the math doesn't work out! Just like I would pay $50-80 extra over a $200 960 for a 290/290X, just like you "future-proofed" with a 5820K over the 4790K, except 290/290X would be faster in almost all games TODAY, not in 3-5 years from now.

15 cents per kWh x 4 hours of gaming x 365 days x 100W difference = $21.90. So it will 5 years to just break even on the electricity costs with a relatively hardcore/'heavy usage' gamer with a 970 vs. 290/290X purchase! I will have sold my R9 290 card way before that. Alternatively, one could get R9 290Xs and buy a new SSD for his/her desktop/laptop for $160 instead of buying 970 SLI which is more or less similar performance. You don't think of things in those terms? Don't you ever need new sports equipment, new smartphone upgrade, new running shoes/shoes. Maybe your kid wanted new console games or a new console, all of a sudden a $400 PS4 will cost you just $240 since well you saved $160 on not spending it on 970 SLI. Am I the only one that looks at things that way?...

Power consumption has now become some 'magical' cop out to justify high prices. $252 Radeon 290X vs. $350 GTX970. Ya, let's pay $100 upfront today to take 5 years to just break even. Does NV think 70% of gamers can't do mathematics? Apparently. :sneaky:

BTW, AMD more or less confirmed no R9 300 series until at least Q2 2015:

"We made progress diversifying our business, ramping design wins and improving our balance sheet this past year despite challenges in our PC business," said Dr. Lisa Su, AMD president and CEO. "Annual Enterprise, Embedded and Semi-Custom segment revenue increased over 50 percent as customer demand for products powered by our high-performance compute and rich visualization solutions was strong. We continue to address channel headwinds in the Computing and Graphics segment and are taking steps to return it to a healthy trajectory beginning in the second quarter of 2015."

For Q1 2015, Advanced Micro Devices expects revenue to decrease 15 percent, plus or minus 3 percent, sequentially.

Read more: http://www.benzinga.com/news/earnin...rts-q4-inline-with-expectations#ixzz3PP9SFAPm
​

Let's try to stay on topic please.

-Rvenger

 

[Headfoot]

Power consumption whining not related to heatsink dissipation potential in enthusiast PC circles = sophistry

 

[DooKey]

I don't understand the pimping of fire sale AMD cards over NV like it's a good thing. AMD cut prices to move product and stay afloat. They aren't pricing their cards to help consumers. The current pricing of AMD cards is an anomaly based upon their poor sales figures and poor financial condition.

Anyway, I hope the next company to buy the ATI IP does something with it. The 3XX series looks good, but it's probably too little, too late to make a difference in the long run. Hopefully I'm wrong because we need 2 strong GPU vendors.

 

[AnandThenMan]

They aren't pricing their cards to help consumers.

Do you really have to state this? What company prices their products to "help" consumers?

 

[Plimogz]

I would also take a $252 R9 290X

Solid deal!

***

I don't know about others, but I welcome a 300W single GPU flagship, assuming the closed loop cooler is quiet and reliable (which seems likely) and the increased power is well spent. (i.e, efficiency doesn't fall off too badly compared to a more mundane air-cooled, 200W model build around the same chip)

I take it that somewhere around 180-250W building a 2-slot air cooled heatsink (of reasonable length, lol) becomes more challenging and expensive (e.g outstanding third party 290X coolers like the Tri-X or Lightning), to say nothing of something built to handle 300+W.

I could see pro and especially HPC users preferring cards optimized for power, hence running bigger chips at lower clocks and volts, money be damned. But gamers? Turn up the clocks and volts, and if the big chip at the heart of the beast scales pretty efficiently up to 300W and you have a viable way to cool the thing -- why not?

 

[digitaldurandal]

(snip 5820 vs 4790 wattage info)

Exar333 Open mouth, insert foot.

The thing everyone is saying is not that the wattage doesn't matter. They are saying that performance is king. In the counter examples you are saying that the wattage is preventing performance and overclocking which in some cases could be true - however if the performance is still better with the heat than another product in that price range, then it does not matter.

Do my CF 290s get hot and prevent extraction of their maximum potential - yes they do. Could have I have gotten more performance for $480? No, even with the heat and power it was still the best deal. Therefor, wattage was irrelevant and will be in most enthusiast situations.

If you are doing custom watercooling to get maximum performance from your equipment that may be another matter entirely because it changes the cost structure. However if we are talking about performance per dollar then in most cases the money spent on watercooling could have been spent somewhere else for performance, with the exception of when using the top of the line GPU etc already.

If we are talking pure performance and cost is not factored then the card using more wattage could still be the best choice as once cooled properly it may be faster than cards that used less wattage and ran cooler to begin with.

So as you can see from my example it is possible when looking at the metrics of performance or performance/$ that a high wattage card is better. Of course the opposite is true as well, but to say that wattage and efficiency are a big deal is simply marketing.

In the end, look at all the data for the GPU or CPU you are considering and compare in that price range. Find the one that gives the best performance in the metrics you desire and purchase. If a product gives good performance and your rig can support it then do not let marketing FUD prevent you from making a smart purchase.

 

[xthetenth]

I don't understand the pimping of fire sale AMD cards over NV like it's a good thing. AMD cut prices to move product and stay afloat. They aren't pricing their cards to help consumers. The current pricing of AMD cards is an anomaly based upon their poor sales figures and poor financial condition.

Anyway, I hope the next company to buy the ATI IP does something with it. The 3XX series looks good, but it's probably too little, too late to make a difference in the long run. Hopefully I'm wrong because we need 2 strong GPU vendors.

If you're coming from the perspective of someone looking to exchange money for components that give good computing performance and to get the most performance for the least money, this is a good opportunity. I hear this perspective is common on tech forums.

From a business point of view, if demand goes up that helps the market so the fire sale pricing doesn't have to be so extreme, and helps us keep two graphics card companies, which is kind of a good thing to have.

I'm a niche case in that a too high TDP inside my case makes my room uncomfortably hot, so I actually have cause to want power efficiency enough to spend twenty-thirty bucks for that alone, and I'm thinking I quite likely overpaid this generation by getting a 970.

 

[Timorous]

With all the conflicting rumours about names, 20nm vs 28nm and so on what is wrong with the following scenario taking place

AMD launching a 380X in the near term to take the crown from the 980 based on a spec sheet like:

28nm
2816 SPs
64 ROPs
176 TMUs
4GB HBM @ 1.25Ghz
GCN 1.2+ (I think they will add better power management features on top of the Tonga improvements but if they call it GCN 1.3 or GCN 2.0 I have no clue).
1.2 Ghz core clock
200 Watts

Then later they launch a 390X to compete with GM200 with a spec sheet like:

20nm
4096 SPs
96 ROPs
256 TMUs
4/8GB HBM @ 2Ghz
GCN 1.2+
?? clock speed
300 Watts
Hybrid cooling

Of course they might call the first card a 390X, or keep the 380X and make the 390X a dual card and release a card similar to the second spec list as a 490X as it is on a new process node but I think in principle this could be what they are trying to do.

 

[Headfoot]

They aren't pricing their cards to help consumers. The current pricing of AMD cards is an anomaly based upon their poor sales figures and poor financial condition.

Who cares? I'm not a stockholder. I just want the most performance at the fewest dollars. I buy used things (gpus included) all the time; I don't lose any sleep over the fact that the second hand sale nets the manufacturer 0 dollars.

GCN 1.2+ (I think they will add better power management features on top of the Tonga improvements but if they call it GCN 1.3 or GCN 2.0 I have no clue)

GCN 1.0, 1.1, 1.2 etc. are not official terminology; just something Anandtech and other tech publishers use to refer to the revisions to the GCN core for the sake of keeping them separate. So what it gets called is whatever the first batch of reviewers decide to start calling it, I suppose

 

[skipsneeky2]

I'm up for a 300w card,especially if performance is good.If my TX650 and its 53 Amps could handle i wouldn't know without proper research but the right price/performance i would be all over it.

I think the power consumption thing has been more of a big deal cause of how efficient Maxwell looks in comparison to the Amd 200 series.It seems people welcome the big beast,it comes then some hippy cards with a set max voltage from Nvidia come out and everyone cares about the environment all over again.

Maybe i could get some of the green guys in here to help me pick up recyclables off the highway and we could go save the earth 2 ways.

 

[Gloomy]

With all the conflicting rumours about names, 20nm vs 28nm and so on what is wrong with the following scenario taking place

AMD launching a 380X in the near term to take the crown from the 980 based on a spec sheet like:

28nm
2816 SPs
64 ROPs
176 TMUs
4GB HBM @ 1.25Ghz
GCN 1.2+ (I think they will add better power management features on top of the Tonga improvements but if they call it GCN 1.3 or GCN 2.0 I have no clue).
1.2 Ghz core clock
200 Watts

Then later they launch a 390X to compete with GM200 with a spec sheet like:

20nm
4096 SPs
96 ROPs
256 TMUs
4/8GB HBM @ 2Ghz
GCN 1.2+
?? clock speed
300 Watts
Hybrid cooling

Of course they might call the first card a 390X, or keep the 380X and make the 390X a dual card and release a card similar to the second spec list as a 490X as it is on a new process node but I think in principle this could be what they are trying to do.

Your specs are ridiculous. 8GB HBM1 at 2GHz would be 2048GB/s or something insane like that

 

[Timorous]

Your specs are ridiculous. 8GB HBM1 at 2GHz would be 2048GB/s or something insane like that

Well sub it for 1.25 Ghz then. The speed of the memory is a total guess. I figured 8GB because a top end card like that will probably target 4K and it would be desirable to have 8GB of ram for that.

 

[Gloomy]

Well sub it for 1.25 Ghz then. The speed of the memory is a total guess. I figured 8GB because a top end card like that will probably target 4K and it would be desirable to have 8GB of ram for that.

That's not even really the problem, the problem is that 8GB would be an 8192bit bus, which I don't think is feasible.

BTW 1.25GHz would turn out to be 1280GB/s which is already faster than the L1 and L2 cache on Hawaii (1TB/s @ 1GHz core clock).

 

[Timorous]

That's not even really the problem, the problem is that 8GB would be an 8192bit bus, which I don't think is feasible.

BTW 1.25GHz would turn out to be 1280GB/s which is already faster than the L1 and L2 cache on Hawaii (1TB/s @ 1GHz core clock).

4GB is not going to be enough for 4k going forward though. I can see a 380X being 4GB for the time being but I do not see GM200 being stuck at 4GB of memory and if AMD do have a 390X to take it on it is going to need 6-8GB to compete at 4k.